void
do_bdr_unregister(void)
{
PGconn *conn = NULL;
ExtensionStatus extension_status = REPMGR_UNKNOWN;
int target_node_id = UNKNOWN_NODE_ID;
t_node_info node_info = T_NODE_INFO_INITIALIZER;
RecordStatus record_status = RECORD_NOT_FOUND;
bool node_record_deleted = false;
PQExpBufferData event_details;
char *dbname;
/* sanity-check configuration for BDR-compatability */
if (config_file_options.replication_type != REPLICATION_TYPE_BDR)
{
log_error(_("cannot run BDR UNREGISTER on a non-BDR node"));
exit(ERR_BAD_CONFIG);
}
dbname = pg_malloc0(MAXLEN);
if (dbname == NULL)
{
log_error(_("unable to allocate memory; terminating."));
exit(ERR_OUT_OF_MEMORY);
}
/* store the database name for future reference */
get_conninfo_value(config_file_options.conninfo, "dbname", dbname);
conn = establish_db_connection(config_file_options.conninfo, true);
if (!is_bdr_db(conn, NULL))
{
log_error(_("database \"%s\" is not BDR-enabled"), dbname);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
extension_status = get_repmgr_extension_status(conn, NULL);
if (extension_status != REPMGR_INSTALLED)
{
log_error(_("repmgr is not installed on database \"%s\""), dbname);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
pfree(dbname);
if (!is_bdr_repmgr(conn))
{
log_error(_("repmgr metadatabase contains records for non-BDR nodes"));
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
initPQExpBuffer(&event_details);
if (runtime_options.node_id != UNKNOWN_NODE_ID)
target_node_id = runtime_options.node_id;
else
target_node_id = config_file_options.node_id;
/* Check node exists and is really a BDR node */
record_status = get_node_record(conn, target_node_id, &node_info);
if (record_status != RECORD_FOUND)
{
log_error(_("no record found for node %i"), target_node_id);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
begin_transaction(conn);
log_debug("unregistering node %i", target_node_id);
node_record_deleted = delete_node_record(conn, target_node_id);
if (node_record_deleted == false)
{
appendPQExpBuffer(&event_details,
"unable to delete node record for node \"%s\" (ID: %i)",
node_info.node_name,
target_node_id);
rollback_transaction(conn);
}
else
{
appendPQExpBuffer(&event_details,
"node record deleted for node \"%s\" (ID: %i)",
node_info.node_name,
target_node_id);
commit_transaction(conn);
}
/* Log the event */
create_event_notification(
conn,
&config_file_options,
config_file_options.node_id,
"bdr_unregister",
true,
event_details.data);
PQfinish(conn);
log_notice(_("bdr node \"%s\" (ID: %i) successfully unregistered"),
node_info.node_name, target_node_id);
termPQExpBuffer(&event_details);
return;
}
bool
reload_config(t_configuration_options *orig_options)
{
PGconn *conn;
t_configuration_options new_options;
bool config_changed = false;
/*
* Re-read the configuration file: repmgr.conf
*/
log_info(_("reloading configuration file and updating repmgr tables\n"));
parse_config(&new_options);
if (new_options.node == -1)
{
log_warning(_("unable to parse new configuration, retaining current configuration\n"));
return false;
}
if (strcmp(new_options.cluster_name, orig_options->cluster_name) != 0)
{
log_warning(_("unable to change cluster name, retaining current configuration\n"));
return false;
}
if (new_options.node != orig_options->node)
{
log_warning(_("unable to change node ID, retaining current configuration\n"));
return false;
}
if (strcmp(new_options.node_name, orig_options->node_name) != 0)
{
log_warning(_("unable to change standby name, keeping current configuration\n"));
return false;
}
if (new_options.failover != MANUAL_FAILOVER && new_options.failover != AUTOMATIC_FAILOVER)
{
log_warning(_("new value for 'failover' must be 'automatic' or 'manual'\n"));
return false;
}
if (new_options.master_response_timeout <= 0)
{
log_warning(_("new value for 'master_response_timeout' must be greater than zero\n"));
return false;
}
if (new_options.reconnect_attempts < 0)
{
log_warning(_("new value for 'reconnect_attempts' must be zero or greater\n"));
return false;
}
if (new_options.reconnect_interval < 0)
{
log_warning(_("new value for 'reconnect_interval' must be zero or greater\n"));
return false;
}
if (strcmp(orig_options->conninfo, new_options.conninfo) != 0)
{
/* Test conninfo string */
conn = establish_db_connection(new_options.conninfo, false);
if (!conn || (PQstatus(conn) != CONNECTION_OK))
{
log_warning(_("'conninfo' string is not valid, retaining current configuration\n"));
return false;
}
PQfinish(conn);
}
/*
* No configuration problems detected - copy any changed values
*
* NB: keep these in the same order as in config.h to make it easier
* to manage them
*/
/* cluster_name */
if (strcmp(orig_options->cluster_name, new_options.cluster_name) != 0)
{
strcpy(orig_options->cluster_name, new_options.cluster_name);
config_changed = true;
}
/* conninfo */
if (strcmp(orig_options->conninfo, new_options.conninfo) != 0)
{
strcpy(orig_options->conninfo, new_options.conninfo);
config_changed = true;
}
/* node */
if (orig_options->node != new_options.node)
{
orig_options->node = new_options.node;
config_changed = true;
}
/* failover */
if (orig_options->failover != new_options.failover)
{
orig_options->failover = new_options.failover;
config_changed = true;
}
/* priority */
if (orig_options->priority != new_options.priority)
{
orig_options->priority = new_options.priority;
config_changed = true;
}
/* node_name */
if (strcmp(orig_options->node_name, new_options.node_name) != 0)
{
strcpy(orig_options->node_name, new_options.node_name);
config_changed = true;
}
/* promote_command */
if (strcmp(orig_options->promote_command, new_options.promote_command) != 0)
{
strcpy(orig_options->promote_command, new_options.promote_command);
config_changed = true;
}
/* follow_command */
if (strcmp(orig_options->follow_command, new_options.follow_command) != 0)
{
strcpy(orig_options->follow_command, new_options.follow_command);
config_changed = true;
}
/*
* XXX These ones can change with a simple SIGHUP?
*
* strcpy (orig_options->loglevel, new_options.loglevel); strcpy
* (orig_options->logfacility, new_options.logfacility);
*
* logger_shutdown(); XXX do we have progname here ? logger_init(progname,
* orig_options.loglevel, orig_options.logfacility);
*/
/* rsync_options */
if (strcmp(orig_options->rsync_options, new_options.rsync_options) != 0)
{
strcpy(orig_options->rsync_options, new_options.rsync_options);
config_changed = true;
}
/* ssh_options */
if (strcmp(orig_options->ssh_options, new_options.ssh_options) != 0)
{
strcpy(orig_options->ssh_options, new_options.ssh_options);
config_changed = true;
}
/* master_response_timeout */
if (orig_options->master_response_timeout != new_options.master_response_timeout)
{
orig_options->master_response_timeout = new_options.master_response_timeout;
config_changed = true;
}
/* reconnect_attempts */
if (orig_options->reconnect_attempts != new_options.reconnect_attempts)
{
orig_options->reconnect_attempts = new_options.reconnect_attempts;
config_changed = true;
}
/* reconnect_interval */
if (orig_options->reconnect_interval != new_options.reconnect_interval)
{
orig_options->reconnect_interval = new_options.reconnect_interval;
config_changed = true;
}
/* pg_ctl_options */
if (strcmp(orig_options->pg_ctl_options, new_options.pg_ctl_options) != 0)
{
strcpy(orig_options->pg_ctl_options, new_options.pg_ctl_options);
config_changed = true;
}
/* pg_basebackup_options */
if (strcmp(orig_options->pg_basebackup_options, new_options.pg_basebackup_options) != 0)
{
strcpy(orig_options->pg_basebackup_options, new_options.pg_basebackup_options);
config_changed = true;
}
/* monitor_interval_secs */
if (orig_options->monitor_interval_secs != new_options.monitor_interval_secs)
{
orig_options->monitor_interval_secs = new_options.monitor_interval_secs;
config_changed = true;
}
/* retry_promote_interval_secs */
if (orig_options->retry_promote_interval_secs != new_options.retry_promote_interval_secs)
{
orig_options->retry_promote_interval_secs = new_options.retry_promote_interval_secs;
config_changed = true;
}
/* use_replication_slots */
if (orig_options->use_replication_slots != new_options.use_replication_slots)
{
orig_options->use_replication_slots = new_options.use_replication_slots;
config_changed = true;
}
if (config_changed == true)
{
log_debug(_("reload_config(): configuration has changed\n"));
}
else
{
log_debug(_("reload_config(): configuration has not changed\n"));
}
return config_changed;
}
PGconn *
get_master_connection(PGconn *standby_conn, char *cluster,
int *master_id, char *master_conninfo_out)
{
PGconn *master_conn = NULL;
PGresult *res1;
PGresult *res2;
char sqlquery[QUERY_STR_LEN];
char master_conninfo_stack[MAXCONNINFO];
char *master_conninfo = &*master_conninfo_stack;
int i,
node_id;
if (master_id != NULL)
{
*master_id = NODE_NOT_FOUND;
}
/* find all nodes belonging to this cluster */
log_info(_("finding node list for cluster '%s'\n"),
cluster);
sqlquery_snprintf(sqlquery,
"SELECT id, conninfo "
" FROM %s.repl_nodes "
" WHERE cluster = '%s' "
" AND type != 'witness' ",
get_repmgr_schema_quoted(standby_conn),
cluster);
res1 = PQexec(standby_conn, sqlquery);
if (PQresultStatus(res1) != PGRES_TUPLES_OK)
{
log_err(_("unable to retrieve node records: %s\n"),
PQerrorMessage(standby_conn));
PQclear(res1);
return NULL;
}
for (i = 0; i < PQntuples(res1); i++)
{
/* initialize with the values of the current node being processed */
node_id = atoi(PQgetvalue(res1, i, 0));
strncpy(master_conninfo, PQgetvalue(res1, i, 1), MAXCONNINFO);
log_info(_("checking role of cluster node '%i'\n"),
node_id);
master_conn = establish_db_connection(master_conninfo, false);
if (PQstatus(master_conn) != CONNECTION_OK)
continue;
/*
* Can't use the is_standby() function here because on error that
* function closes the connection passed and exits. This still needs
* to close master_conn first.
*/
res2 = PQexec(master_conn, "SELECT pg_catalog.pg_is_in_recovery()");
if (PQresultStatus(res2) != PGRES_TUPLES_OK)
{
log_err(_("unable to retrieve recovery state from this node: %s\n"),
PQerrorMessage(master_conn));
PQclear(res2);
PQfinish(master_conn);
continue;
}
/* if false, this is the master */
if (strcmp(PQgetvalue(res2, 0, 0), "f") == 0)
{
PQclear(res2);
PQclear(res1);
log_debug(_("get_master_connection(): current master node is %i\n"), node_id);
if (master_id != NULL)
{
*master_id = node_id;
}
return master_conn;
}
else
{
/* if it is a standby, clear info */
PQclear(res2);
PQfinish(master_conn);
}
}
/*
* If we finish this loop without finding a master then we doesn't have
* the info or the master has failed (or we reached max_connections or
* superuser_reserved_connections, anything else I'm missing?).
*
* Probably we will need to check the error to know if we need to start
* failover procedure or just fix some situation on the standby.
*/
PQclear(res1);
return NULL;
}
/*
* do_bdr_register()
*
* As each BDR node is its own primary, registering a BDR node
* will create the repmgr metadata schema if necessary.
*/
void
do_bdr_register(void)
{
PGconn *conn = NULL;
BdrNodeInfoList bdr_nodes = T_BDR_NODE_INFO_LIST_INITIALIZER;
ExtensionStatus extension_status = REPMGR_UNKNOWN;
t_node_info node_info = T_NODE_INFO_INITIALIZER;
RecordStatus record_status = RECORD_NOT_FOUND;
PQExpBufferData event_details;
bool success = true;
char *dbname = NULL;
/* sanity-check configuration for BDR-compatability */
if (config_file_options.replication_type != REPLICATION_TYPE_BDR)
{
log_error(_("cannot run BDR REGISTER on a non-BDR node"));
exit(ERR_BAD_CONFIG);
}
dbname = pg_malloc0(MAXLEN);
if (dbname == NULL)
{
log_error(_("unable to allocate memory; terminating."));
exit(ERR_OUT_OF_MEMORY);
}
/* store the database name for future reference */
get_conninfo_value(config_file_options.conninfo, "dbname", dbname);
conn = establish_db_connection(config_file_options.conninfo, true);
if (!is_bdr_db(conn, NULL))
{
log_error(_("database \"%s\" is not BDR-enabled"), dbname);
log_hint(_("when using repmgr with BDR, the repmgr schema must be stored in the BDR database"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
/* Check that there are at most 2 BDR nodes */
get_all_bdr_node_records(conn, &bdr_nodes);
if (bdr_nodes.node_count == 0)
{
log_error(_("database \"%s\" is BDR-enabled but no BDR nodes were found"), dbname);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
/* BDR 2 implementation is for 2 nodes only */
if (get_bdr_version_num() < 3 && bdr_nodes.node_count > 2)
{
log_error(_("repmgr can only support BDR 2.x clusters with 2 nodes"));
log_detail(_("this BDR cluster has %i nodes"), bdr_nodes.node_count);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
if (get_bdr_version_num() > 2)
{
log_error(_("\"repmgr bdr register\" is for BDR 2.x only"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
/* check for a matching BDR node */
{
PQExpBufferData bdr_local_node_name;
bool node_match = false;
initPQExpBuffer(&bdr_local_node_name);
node_match = bdr_node_name_matches(conn, config_file_options.node_name, &bdr_local_node_name);
if (node_match == false)
{
if (strlen(bdr_local_node_name.data))
{
log_error(_("local node BDR node name is \"%s\", expected: \"%s\""),
bdr_local_node_name.data,
config_file_options.node_name);
log_hint(_("\"node_name\" in repmgr.conf must match \"node_name\" in bdr.bdr_nodes"));
}
else
{
log_error(_("local node does not report BDR node name"));
log_hint(_("ensure this is an active BDR node"));
}
PQfinish(conn);
pfree(dbname);
termPQExpBuffer(&bdr_local_node_name);
exit(ERR_BAD_CONFIG);
}
termPQExpBuffer(&bdr_local_node_name);
}
/* check whether repmgr extension exists, and there are no non-BDR nodes registered */
extension_status = get_repmgr_extension_status(conn, NULL);
if (extension_status == REPMGR_UNKNOWN)
{
log_error(_("unable to determine status of \"repmgr\" extension in database \"%s\""),
dbname);
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
if (extension_status == REPMGR_UNAVAILABLE)
{
log_error(_("\"repmgr\" extension is not available"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
if (extension_status == REPMGR_INSTALLED)
{
if (!is_bdr_repmgr(conn))
{
log_error(_("repmgr metadatabase contains records for non-BDR nodes"));
PQfinish(conn);
pfree(dbname);
exit(ERR_BAD_CONFIG);
}
}
else
{
log_debug("creating repmgr extension in database \"%s\"", dbname);
begin_transaction(conn);
if (!create_repmgr_extension(conn))
{
log_error(_("unable to create repmgr extension - see preceding error message(s); aborting"));
rollback_transaction(conn);
pfree(dbname);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
commit_transaction(conn);
}
pfree(dbname);
if (bdr_node_has_repmgr_set(conn, config_file_options.node_name) == false)
{
log_debug("bdr_node_has_repmgr_set() = false");
bdr_node_set_repmgr_set(conn, config_file_options.node_name);
}
/*
* before adding the extension tables to the replication set, if any other
* BDR nodes exist, populate repmgr.nodes with a copy of existing entries
*
* currently we won't copy the contents of any other tables
*
*/
{
NodeInfoList local_node_records = T_NODE_INFO_LIST_INITIALIZER;
(void) get_all_node_records(conn, &local_node_records);
if (local_node_records.node_count == 0)
{
BdrNodeInfoList bdr_nodes = T_BDR_NODE_INFO_LIST_INITIALIZER;
BdrNodeInfoListCell *bdr_cell = NULL;
get_all_bdr_node_records(conn, &bdr_nodes);
if (bdr_nodes.node_count == 0)
{
log_error(_("unable to retrieve any BDR node records"));
log_detail("%s", PQerrorMessage(conn));
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
for (bdr_cell = bdr_nodes.head; bdr_cell; bdr_cell = bdr_cell->next)
{
PGconn *bdr_node_conn = NULL;
NodeInfoList existing_nodes = T_NODE_INFO_LIST_INITIALIZER;
NodeInfoListCell *cell = NULL;
ExtensionStatus other_node_extension_status = REPMGR_UNKNOWN;
/* skip the local node */
if (strncmp(node_info.node_name, bdr_cell->node_info->node_name, sizeof(node_info.node_name)) == 0)
{
continue;
}
log_debug("connecting to BDR node \"%s\" (conninfo: \"%s\")",
bdr_cell->node_info->node_name,
bdr_cell->node_info->node_local_dsn);
bdr_node_conn = establish_db_connection_quiet(bdr_cell->node_info->node_local_dsn);
if (PQstatus(bdr_node_conn) != CONNECTION_OK)
{
continue;
}
/* check repmgr schema exists, skip if not */
other_node_extension_status = get_repmgr_extension_status(bdr_node_conn, NULL);
if (other_node_extension_status != REPMGR_INSTALLED)
{
continue;
}
(void) get_all_node_records(bdr_node_conn, &existing_nodes);
for (cell = existing_nodes.head; cell; cell = cell->next)
{
log_debug("creating record for node \"%s\" (ID: %i)",
cell->node_info->node_name, cell->node_info->node_id);
create_node_record(conn, "bdr register", cell->node_info);
}
PQfinish(bdr_node_conn);
break;
}
}
}
/* Add the repmgr extension tables to a replication set */
if (get_bdr_version_num() < 3)
{
add_extension_tables_to_bdr_replication_set(conn);
}
else
{
/* this is the only table we need to replicate */
char *replication_set = get_default_bdr_replication_set(conn);
/*
* this probably won't happen, but we need to be sure we're using
* the replication set metadata correctly...
*/
if (conn == NULL)
{
log_error(_("unable to retrieve default BDR replication set"));
log_hint(_("see preceding messages"));
log_debug("check query in get_default_bdr_replication_set()");
exit(ERR_BAD_CONFIG);
}
if (is_table_in_bdr_replication_set(conn, "nodes", replication_set) == false)
{
add_table_to_bdr_replication_set(conn, "nodes", replication_set);
}
pfree(replication_set);
}
initPQExpBuffer(&event_details);
begin_transaction(conn);
/*
* we'll check if a record exists (even if the schema was just created),
* as there's a faint chance of a race condition
*/
record_status = get_node_record(conn, config_file_options.node_id, &node_info);
/* Update internal node record */
node_info.type = BDR;
node_info.node_id = config_file_options.node_id;
node_info.upstream_node_id = NO_UPSTREAM_NODE;
node_info.active = true;
node_info.priority = config_file_options.priority;
strncpy(node_info.node_name, config_file_options.node_name, sizeof(node_info.node_name));
strncpy(node_info.location, config_file_options.location, sizeof(node_info.location));
strncpy(node_info.conninfo, config_file_options.conninfo, sizeof(node_info.conninfo));
if (record_status == RECORD_FOUND)
{
bool node_updated = false;
/*
* At this point we will have established there are no non-BDR
* records, so no need to verify the node type
*/
if (!runtime_options.force)
{
log_error(_("this node is already registered"));
log_hint(_("use -F/--force to overwrite the existing node record"));
rollback_transaction(conn);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
/*
* don't permit changing the node name - this must match the BDR node
* name set when the node was registered.
*/
if (strncmp(node_info.node_name, config_file_options.node_name, sizeof(node_info.node_name)) != 0)
{
log_error(_("a record for node %i is already registered with node_name \"%s\""),
config_file_options.node_id, node_info.node_name);
log_hint(_("node_name configured in repmgr.conf is \"%s\""), config_file_options.node_name);
rollback_transaction(conn);
PQfinish(conn);
exit(ERR_BAD_CONFIG);
}
node_updated = update_node_record(conn, "bdr register", &node_info);
if (node_updated == true)
{
appendPQExpBuffer(&event_details, _("node record updated for node \"%s\" (%i)"),
config_file_options.node_name, config_file_options.node_id);
log_verbose(LOG_NOTICE, "%s", event_details.data);
}
else
{
success = false;
}
}
else
{
/* create new node record */
bool node_created = create_node_record(conn, "bdr register", &node_info);
if (node_created == true)
{
appendPQExpBuffer(&event_details,
_("node record created for node \"%s\" (ID: %i)"),
config_file_options.node_name, config_file_options.node_id);
log_notice("%s", event_details.data);
}
else
{
success = false;
}
}
if (success == false)
{
rollback_transaction(conn);
PQfinish(conn);
exit(ERR_DB_QUERY);
}
commit_transaction(conn);
/* Log the event */
create_event_notification(
conn,
&config_file_options,
config_file_options.node_id,
"bdr_register",
true,
event_details.data);
termPQExpBuffer(&event_details);
PQfinish(conn);
log_notice(_("BDR node %i registered (conninfo: %s)"),
config_file_options.node_id, config_file_options.conninfo);
return;
}
/*
* get_upstream_connection()
*
* Returns connection to node's upstream node
*
* NOTE: will attempt to connect even if node is marked as inactive
*/
PGconn *
get_upstream_connection(PGconn *standby_conn, char *cluster, int node_id,
int *upstream_node_id_ptr, char *upstream_conninfo_out)
{
PGconn *upstream_conn = NULL;
PGresult *res;
char sqlquery[QUERY_STR_LEN];
char upstream_conninfo_stack[MAXCONNINFO];
char *upstream_conninfo = &*upstream_conninfo_stack;
/*
* If the caller wanted to get a copy of the connection info string, sub
* out the local stack pointer for the pointer passed by the caller.
*/
if (upstream_conninfo_out != NULL)
upstream_conninfo = upstream_conninfo_out;
sqlquery_snprintf(sqlquery,
" SELECT un.conninfo, un.name, un.id "
" FROM %s.repl_nodes un "
"INNER JOIN %s.repl_nodes n "
" ON (un.id = n.upstream_node_id AND un.cluster = n.cluster)"
" WHERE n.cluster = '%s' "
" AND n.id = %i ",
get_repmgr_schema_quoted(standby_conn),
get_repmgr_schema_quoted(standby_conn),
cluster,
node_id);
log_debug("get_upstream_connection(): %s\n", sqlquery);
res = PQexec(standby_conn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err(_("unable to get conninfo for upstream server: %s\n"),
PQerrorMessage(standby_conn));
PQclear(res);
return NULL;
}
if (!PQntuples(res))
{
log_notice(_("no record found for upstream server"));
PQclear(res);
return NULL;
}
strncpy(upstream_conninfo, PQgetvalue(res, 0, 0), MAXCONNINFO);
if (upstream_node_id_ptr != NULL)
*upstream_node_id_ptr = atoi(PQgetvalue(res, 0, 1));
PQclear(res);
log_debug("conninfo is: '%s'\n", upstream_conninfo);
upstream_conn = establish_db_connection(upstream_conninfo, false);
if (PQstatus(upstream_conn) != CONNECTION_OK)
{
log_err(_("unable to connect to upstream node: %s\n"),
PQerrorMessage(upstream_conn));
return NULL;
}
return upstream_conn;
}
static void
do_failover(void)
{
PGresult *res;
char sqlquery[QUERY_STR_LEN];
int total_nodes = 0;
int visible_nodes = 0;
int ready_nodes = 0;
bool find_best = false;
int i;
int r;
uint32 uxlogid;
uint32 uxrecoff;
XLogRecPtr xlog_recptr;
char last_wal_standby_applied[MAXLEN];
PGconn *node_conn = NULL;
/*
* will get info about until 50 nodes, which seems to be large enough for
* most scenarios
*/
t_node_info nodes[50];
/* initialize to keep compiler quiet */
t_node_info best_candidate = {-1, "", InvalidXLogRecPtr, false, false, false};
/* get a list of standby nodes, including myself */
sprintf(sqlquery, "SELECT id, conninfo, witness "
" FROM %s.repl_nodes "
" WHERE cluster = '%s' "
" ORDER BY priority, id ",
repmgr_schema, local_options.cluster_name);
res = PQexec(my_local_conn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err(_("Can't get nodes' info: %s\n"), PQerrorMessage(my_local_conn));
PQclear(res);
terminate(ERR_DB_QUERY);
}
/*
* total nodes that are registered
*/
total_nodes = PQntuples(res);
log_debug(_("%s: there are %d nodes registered\n"), progname, total_nodes);
/*
* Build an array with the nodes and indicate which ones are visible and
* ready
*/
for (i = 0; i < total_nodes; i++)
{
nodes[i].node_id = atoi(PQgetvalue(res, i, 0));
strncpy(nodes[i].conninfo_str, PQgetvalue(res, i, 1), MAXLEN);
nodes[i].is_witness = (strcmp(PQgetvalue(res, i, 2), "t") == 0) ? true : false;
/*
* Initialize on false so if we can't reach this node we know that
* later
*/
nodes[i].is_visible = false;
nodes[i].is_ready = false;
XLAssignValue(nodes[i].xlog_location, 0, 0);
log_debug(_("%s: node=%d conninfo=\"%s\" witness=%s\n"),
progname, nodes[i].node_id, nodes[i].conninfo_str,
(nodes[i].is_witness) ? "true" : "false");
node_conn = establish_db_connection(nodes[i].conninfo_str, false);
/* if we can't see the node just skip it */
if (PQstatus(node_conn) != CONNECTION_OK)
{
if (node_conn != NULL)
PQfinish(node_conn);
continue;
}
visible_nodes++;
nodes[i].is_visible = true;
PQfinish(node_conn);
}
PQclear(res);
log_debug(_("Total nodes counted: registered=%d, visible=%d\n"),
total_nodes, visible_nodes);
/*
* am i on the group that should keep alive? if i see less than half of
* total_nodes then i should do nothing
*/
if (visible_nodes < (total_nodes / 2.0))
{
log_err(_("Can't reach most of the nodes.\n"
"Let the other standby servers decide which one will be the primary.\n"
"Manual action will be needed to readd this node to the cluster.\n"));
terminate(ERR_FAILOVER_FAIL);
}
/* Query all the nodes to determine which ones are ready */
for (i = 0; i < total_nodes; i++)
{
/* if the node is not visible, skip it */
if (!nodes[i].is_visible)
continue;
if (nodes[i].is_witness)
continue;
node_conn = establish_db_connection(nodes[i].conninfo_str, false);
/*
* XXX This shouldn't happen, if this happens it means this is a major
* problem maybe network outages? anyway, is better for a human to
* react
*/
if (PQstatus(node_conn) != CONNECTION_OK)
{
log_err(_("It seems new problems are arising, manual intervention is needed\n"));
terminate(ERR_FAILOVER_FAIL);
}
uxlogid = 0;
uxrecoff = 0;
sqlquery_snprintf(sqlquery, "SELECT pg_last_xlog_receive_location()");
res = PQexec(node_conn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_info(_("Can't get node's last standby location: %s\n"),
PQerrorMessage(node_conn));
log_info(_("Connection details: %s\n"), nodes[i].conninfo_str);
PQclear(res);
PQfinish(node_conn);
terminate(ERR_FAILOVER_FAIL);
}
if (sscanf(PQgetvalue(res, 0, 0), "%X/%X", &uxlogid, &uxrecoff) != 2)
log_info(_("could not parse transaction log location \"%s\"\n"),
PQgetvalue(res, 0, 0));
log_debug("XLog position of node %d: log id=%u (%X), offset=%u (%X)\n",
nodes[i].node_id, uxlogid, uxlogid, uxrecoff, uxrecoff);
/* If position is 0/0, error */
if (uxlogid == 0 && uxrecoff == 0)
{
PQclear(res);
PQfinish(node_conn);
log_info(_("InvalidXLogRecPtr detected in a standby\n"));
terminate(ERR_FAILOVER_FAIL);
}
XLAssignValue(nodes[i].xlog_location, uxlogid, uxrecoff);
PQclear(res);
PQfinish(node_conn);
}
/* last we get info about this node, and update shared memory */
sprintf(sqlquery, "SELECT pg_last_xlog_receive_location()");
res = PQexec(my_local_conn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err(_("PQexec failed: %s.\nReport an invalid value to not be "
" considered as new primary and exit.\n"),
PQerrorMessage(my_local_conn));
PQclear(res);
sprintf(last_wal_standby_applied, "'%X/%X'", 0, 0);
update_shared_memory(last_wal_standby_applied);
terminate(ERR_DB_QUERY);
}
/* write last location in shared memory */
update_shared_memory(PQgetvalue(res, 0, 0));
PQclear(res);
for (i = 0; i < total_nodes; i++)
{
while (!nodes[i].is_ready)
{
/*
* the witness will always be masked as ready if it's still not
* marked that way and avoid a useless query
*/
if (nodes[i].is_witness)
{
if (!nodes[i].is_ready)
{
nodes[i].is_ready = true;
ready_nodes++;
}
break;
}
/* if the node is not visible, skip it */
if (!nodes[i].is_visible)
break;
/* if the node is ready there is nothing to check, skip it too */
if (nodes[i].is_ready)
break;
node_conn = establish_db_connection(nodes[i].conninfo_str, false);
/*
* XXX This shouldn't happen, if this happens it means this is a
* major problem maybe network outages? anyway, is better for a
* human to react
*/
if (PQstatus(node_conn) != CONNECTION_OK)
{
/* XXX */
log_info(_("At this point, it could be some race conditions "
"that are acceptable, assume the node is restarting "
"and starting failover procedure\n"));
break;
}
uxlogid = 0;
uxrecoff = 0;
sqlquery_snprintf(sqlquery, "SELECT %s.repmgr_get_last_standby_location()",
repmgr_schema);
res = PQexec(node_conn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err(_("PQexec failed: %s.\nReport an invalid value to not"
"be considered as new primary and exit.\n"),
PQerrorMessage(node_conn));
PQclear(res);
PQfinish(node_conn);
terminate(ERR_DB_QUERY);
}
if (sscanf(PQgetvalue(res, 0, 0), "%X/%X", &uxlogid, &uxrecoff) != 2)
{
log_info(_("could not parse transaction log location \"%s\"\n"),
PQgetvalue(res, 0, 0));
/* we can't do anything but fail at this point... */
if (*PQgetvalue(res, 0, 0) == '\0')
{
log_crit("Whoops, seems as if shared_preload_libraries=repmgr_funcs is not set!\n");
exit(ERR_BAD_CONFIG);
}
}
PQclear(res);
PQfinish(node_conn);
/* If position is 0/0, keep checking */
if (uxlogid == 0 && uxrecoff == 0)
continue;
XLAssignValue(xlog_recptr, uxlogid, uxrecoff);
if (XLByteLT(nodes[i].xlog_location, xlog_recptr))
{
XLAssignValue(nodes[i].xlog_location, uxlogid, uxrecoff);
}
log_debug("Last XLog position of node %d: log id=%u (%X), offset=%u (%X)\n",
nodes[i].node_id, uxlogid, uxlogid,
uxrecoff, uxrecoff);
ready_nodes++;
nodes[i].is_ready = true;
}
}
/* Close the connection to this server */
PQfinish(my_local_conn);
my_local_conn = NULL;
/*
* determine which one is the best candidate to promote to primary
*/
for (i = 0; i < total_nodes; i++)
{
/* witness is never a good candidate */
if (nodes[i].is_witness)
continue;
if (!nodes[i].is_ready || !nodes[i].is_visible)
continue;
if (!find_best)
{
/*
* start with the first ready node, and then move on to the next
* one
*/
best_candidate.node_id = nodes[i].node_id;
XLAssign(best_candidate.xlog_location, nodes[i].xlog_location);
best_candidate.is_ready = nodes[i].is_ready;
best_candidate.is_witness = nodes[i].is_witness;
find_best = true;
}
/* we use the macros provided by xlogdefs.h to compare XLogRecPtr */
/*
* Nodes are retrieved ordered by priority, so if the current best
* candidate is lower than the next node's wal location then assign
* next node as the new best candidate.
*/
if (XLByteLT(best_candidate.xlog_location, nodes[i].xlog_location))
{
best_candidate.node_id = nodes[i].node_id;
XLAssign(best_candidate.xlog_location, nodes[i].xlog_location);
best_candidate.is_ready = nodes[i].is_ready;
best_candidate.is_witness = nodes[i].is_witness;
}
}
/* once we know who is the best candidate, promote it */
if (find_best && (best_candidate.node_id == local_options.node))
{
if (best_candidate.is_witness)
{
log_err(_("%s: Node selected as new master is a witness. Can't be promoted.\n"),
progname);
terminate(ERR_FAILOVER_FAIL);
}
/* wait */
sleep(5);
if (verbose)
log_info(_("%s: This node is the best candidate to be the new primary, promoting...\n"),
progname);
log_debug(_("promote command is: \"%s\"\n"),
local_options.promote_command);
if (log_type == REPMGR_STDERR && *local_options.logfile)
{
fflush(stderr);
}
r = system(local_options.promote_command);
if (r != 0)
{
log_err(_("%s: promote command failed. You could check and try it manually.\n"),
progname);
terminate(ERR_BAD_CONFIG);
}
}
else if (find_best)
{
/* wait */
sleep(10);
if (verbose)
log_info(_("%s: Node %d is the best candidate to be the new primary, we should follow it...\n"),
progname, best_candidate.node_id);
log_debug(_("follow command is: \"%s\"\n"), local_options.follow_command);
/*
* New Primary need some time to be promoted. The follow command
* should take care of that.
*/
if (log_type == REPMGR_STDERR && *local_options.logfile)
{
fflush(stderr);
}
r = system(local_options.follow_command);
if (r != 0)
{
log_err(_("%s: follow command failed. You could check and try it manually.\n"),
progname);
terminate(ERR_BAD_CONFIG);
}
}
else
{
log_err(_("%s: Did not find candidates. You should check and try manually.\n"),
progname);
terminate(ERR_FAILOVER_FAIL);
}
/* to force it to re-calculate mode and master node */
failover_done = true;
/* and reconnect to the local database */
my_local_conn = establish_db_connection(local_options.conninfo, true);
}
int
main(int argc, char **argv)
{
static struct option long_options[] =
{
{"config-file", required_argument, NULL, 'f'},
{"verbose", no_argument, NULL, 'v'},
{"monitoring-history", no_argument, NULL, 'm'},
{"daemonize", no_argument, NULL, 'd'},
{"pid-file", required_argument, NULL, 'p'},
{NULL, 0, NULL, 0}
};
int optindex;
int c,
ret;
bool daemonize = false;
FILE *fd;
char standby_version[MAXVERSIONSTR],
*ret_ver;
progname = get_progname(argv[0]);
if (argc > 1)
{
if (strcmp(argv[1], "--help") == 0 || strcmp(argv[1], "-?") == 0)
{
help(progname);
exit(SUCCESS);
}
if (strcmp(argv[1], "--version") == 0 || strcmp(argv[1], "-V") == 0)
{
printf("%s %s (PostgreSQL %s)\n", progname, REPMGR_VERSION, PG_VERSION);
exit(SUCCESS);
}
}
while ((c = getopt_long(argc, argv, "f:v:mdp:", long_options, &optindex)) != -1)
{
switch (c)
{
case 'f':
config_file = optarg;
break;
case 'v':
verbose = true;
break;
case 'm':
monitoring_history = true;
break;
case 'd':
daemonize = true;
break;
case 'p':
pid_file = optarg;
break;
default:
usage();
exit(ERR_BAD_CONFIG);
}
}
if (daemonize)
{
do_daemonize();
}
if (pid_file)
{
check_and_create_pid_file(pid_file);
}
#ifndef WIN32
setup_event_handlers();
#endif
/*
* Read the configuration file: repmgr.conf
*/
parse_config(config_file, &local_options);
if (local_options.node == -1)
{
log_err(_("Node information is missing. "
"Check the configuration file, or provide one if you have not done so.\n"));
terminate(ERR_BAD_CONFIG);
}
fd = freopen("/dev/null", "r", stdin);
if (fd == NULL)
{
fprintf(stderr, "error reopening stdin to '/dev/null': %s",
strerror(errno));
}
fd = freopen("/dev/null", "w", stdout);
if (fd == NULL)
{
fprintf(stderr, "error reopening stdout to '/dev/null': %s",
strerror(errno));
}
logger_init(&local_options, progname, local_options.loglevel,
local_options.logfacility);
if (verbose)
logger_min_verbose(LOG_INFO);
if (log_type == REPMGR_SYSLOG)
{
fd = freopen("/dev/null", "w", stderr);
if (fd == NULL)
{
fprintf(stderr, "error reopening stderr to '/dev/null': %s",
strerror(errno));
}
}
xsnprintf(repmgr_schema, MAXLEN, "%s%s", DEFAULT_REPMGR_SCHEMA_PREFIX,
local_options.cluster_name);
log_info(_("%s Connecting to database '%s'\n"), progname,
local_options.conninfo);
my_local_conn = establish_db_connection(local_options.conninfo, true);
/* should be v9 or better */
log_info(_("%s Connected to database, checking its state\n"), progname);
ret_ver = pg_version(my_local_conn, standby_version);
if (ret_ver == NULL || strcmp(standby_version, "") == 0)
{
if (ret_ver != NULL)
log_err(_("%s needs standby to be PostgreSQL 9.0 or better\n"),
progname);
terminate(ERR_BAD_CONFIG);
}
/*
* MAIN LOOP This loops cicles once per failover and at startup
* Requisites: - my_local_conn needs to be already setted with an active
* connection - no master connection
*/
do
{
/*
* Set my server mode, establish a connection to primary and start
* monitor
*/
ret = is_witness(my_local_conn, repmgr_schema,
local_options.cluster_name, local_options.node);
if (ret == 1)
my_local_mode = WITNESS_MODE;
else if (ret == 0)
{
ret = is_standby(my_local_conn);
if (ret == 1)
my_local_mode = STANDBY_MODE;
else if (ret == 0) /* is the master */
my_local_mode = PRIMARY_MODE;
}
/*
* XXX we did this before changing is_standby() to return int; we
* should not exit at this point, but for now we do until we have a
* better strategy
*/
if (ret == -1)
terminate(1);
switch (my_local_mode)
{
case PRIMARY_MODE:
primary_options.node = local_options.node;
strncpy(primary_options.conninfo, local_options.conninfo,
MAXLEN);
primary_conn = my_local_conn;
check_cluster_configuration(my_local_conn);
check_node_configuration();
if (reload_config(config_file, &local_options))
{
PQfinish(my_local_conn);
my_local_conn = establish_db_connection(local_options.conninfo, true);
primary_conn = my_local_conn;
update_registration();
}
log_info(_("%s Starting continuous primary connection check\n"),
progname);
/*
* Check that primary is still alive, and standbies are
* sending info
*/
/*
* Every local_options.monitor_interval_secs seconds, do
* master checks XXX Check that standbies are sending info
*/
do
{
if (check_connection(primary_conn, "master"))
{
/*
* CheckActiveStandbiesConnections();
* CheckInactiveStandbies();
*/
sleep(local_options.monitor_interval_secs);
}
else
{
/*
* XXX May we do something more verbose ?
*/
terminate(1);
}
if (got_SIGHUP)
{
/*
* if we can reload, then could need to change
* my_local_conn
*/
if (reload_config(config_file, &local_options))
{
PQfinish(my_local_conn);
my_local_conn = establish_db_connection(local_options.conninfo, true);
primary_conn = my_local_conn;
if (*local_options.logfile)
{
FILE *fd;
fd = freopen(local_options.logfile, "a", stderr);
if (fd == NULL)
{
fprintf(stderr, "error reopening stderr to '%s': %s",
local_options.logfile, strerror(errno));
}
}
update_registration();
}
got_SIGHUP = false;
}
} while (!failover_done);
break;
case WITNESS_MODE:
case STANDBY_MODE:
/* I need the id of the primary as well as a connection to it */
log_info(_("%s Connecting to primary for cluster '%s'\n"),
progname, local_options.cluster_name);
primary_conn = get_master_connection(my_local_conn, repmgr_schema,
local_options.cluster_name,
&primary_options.node, NULL);
if (primary_conn == NULL)
{
terminate(ERR_BAD_CONFIG);
}
check_cluster_configuration(my_local_conn);
check_node_configuration();
if (reload_config(config_file, &local_options))
{
PQfinish(my_local_conn);
my_local_conn = establish_db_connection(local_options.conninfo, true);
update_registration();
}
/*
* Every local_options.monitor_interval_secs seconds, do
* checks
*/
if (my_local_mode == WITNESS_MODE)
{
log_info(_("%s Starting continuous witness node monitoring\n"),
progname);
}
else if (my_local_mode == STANDBY_MODE)
{
log_info(_("%s Starting continuous standby node monitoring\n"),
progname);
}
do
{
if (my_local_mode == WITNESS_MODE)
witness_monitor();
else if (my_local_mode == STANDBY_MODE)
standby_monitor();
sleep(local_options.monitor_interval_secs);
if (got_SIGHUP)
{
/*
* if we can reload, then could need to change
* my_local_conn
*/
if (reload_config(config_file, &local_options))
{
PQfinish(my_local_conn);
my_local_conn = establish_db_connection(local_options.conninfo, true);
update_registration();
}
got_SIGHUP = false;
}
} while (!failover_done);
break;
default:
log_err(_("%s: Unrecognized mode for node %d\n"), progname,
local_options.node);
}
failover_done = false;
} while (true);
/* close the connection to the database and cleanup */
close_connections();
/* Shuts down logging system */
logger_shutdown();
return 0;
}
/*
* reload_config()
*
* This is only called by repmgrd after receiving a SIGHUP or when a monitoring
* loop is started up; it therefore only needs to reload options required
* by repmgrd, which are as follows:
*
* changeable options:
* - failover
* - follow_command
* - logfacility
* - logfile
* - loglevel
* - master_response_timeout
* - monitor_interval_secs
* - priority
* - promote_command
* - reconnect_attempts
* - reconnect_interval
* - retry_promote_interval_secs
* - witness_repl_nodes_sync_interval_secs
*
* non-changeable options:
* - cluster_name
* - conninfo
* - node
* - node_name
*
* extract with something like:
* grep local_options\\. repmgrd.c | perl -n -e '/local_options\.([\w_]+)/ && print qq|$1\n|;' | sort | uniq
*/
bool
reload_config(t_configuration_options *orig_options)
{
PGconn *conn;
t_configuration_options new_options = T_CONFIGURATION_OPTIONS_INITIALIZER;
bool config_changed = false;
bool log_config_changed = false;
static ItemList config_errors = { NULL, NULL };
/*
* Re-read the configuration file: repmgr.conf
*/
log_info(_("reloading configuration file\n"));
_parse_config(&new_options, &config_errors);
if (config_errors.head != NULL)
{
/* XXX dump errors to log */
log_warning(_("unable to parse new configuration, retaining current configuration\n"));
return false;
}
/* The following options cannot be changed */
if (strcmp(new_options.cluster_name, orig_options->cluster_name) != 0)
{
log_warning(_("cluster_name cannot be changed, retaining current configuration\n"));
return false;
}
if (new_options.node != orig_options->node)
{
log_warning(_("node ID cannot be changed, retaining current configuration\n"));
return false;
}
if (strcmp(new_options.node_name, orig_options->node_name) != 0)
{
log_warning(_("node_name cannot be changed, keeping current configuration\n"));
return false;
}
if (strcmp(orig_options->conninfo, new_options.conninfo) != 0)
{
/* Test conninfo string works*/
conn = establish_db_connection(new_options.conninfo, false);
if (!conn || (PQstatus(conn) != CONNECTION_OK))
{
log_warning(_("'conninfo' string is not valid, retaining current configuration\n"));
return false;
}
PQfinish(conn);
}
/*
* No configuration problems detected - copy any changed values
*
* NB: keep these in the same order as in config.h to make it easier
* to manage them
*/
/* failover */
if (orig_options->failover != new_options.failover)
{
orig_options->failover = new_options.failover;
config_changed = true;
}
/* follow_command */
if (strcmp(orig_options->follow_command, new_options.follow_command) != 0)
{
strcpy(orig_options->follow_command, new_options.follow_command);
config_changed = true;
}
/* master_response_timeout */
if (orig_options->master_response_timeout != new_options.master_response_timeout)
{
orig_options->master_response_timeout = new_options.master_response_timeout;
config_changed = true;
}
/* monitor_interval_secs */
if (orig_options->monitor_interval_secs != new_options.monitor_interval_secs)
{
orig_options->monitor_interval_secs = new_options.monitor_interval_secs;
config_changed = true;
}
/* priority */
if (orig_options->priority != new_options.priority)
{
orig_options->priority = new_options.priority;
config_changed = true;
}
/* promote_command */
if (strcmp(orig_options->promote_command, new_options.promote_command) != 0)
{
strcpy(orig_options->promote_command, new_options.promote_command);
config_changed = true;
}
/* reconnect_attempts */
if (orig_options->reconnect_attempts != new_options.reconnect_attempts)
{
orig_options->reconnect_attempts = new_options.reconnect_attempts;
config_changed = true;
}
/* reconnect_interval */
if (orig_options->reconnect_interval != new_options.reconnect_interval)
{
orig_options->reconnect_interval = new_options.reconnect_interval;
config_changed = true;
}
/* retry_promote_interval_secs */
if (orig_options->retry_promote_interval_secs != new_options.retry_promote_interval_secs)
{
orig_options->retry_promote_interval_secs = new_options.retry_promote_interval_secs;
config_changed = true;
}
/* witness_repl_nodes_sync_interval_secs */
if (orig_options->witness_repl_nodes_sync_interval_secs != new_options.witness_repl_nodes_sync_interval_secs)
{
orig_options->witness_repl_nodes_sync_interval_secs = new_options.witness_repl_nodes_sync_interval_secs;
config_changed = true;
}
/*
* Handle changes to logging configuration
*/
if (strcmp(orig_options->logfacility, new_options.logfacility) != 0)
{
strcpy(orig_options->logfacility, new_options.logfacility);
log_config_changed = true;
}
if (strcmp(orig_options->logfile, new_options.logfile) != 0)
{
strcpy(orig_options->logfile, new_options.logfile);
log_config_changed = true;
}
if (strcmp(orig_options->loglevel, new_options.loglevel) != 0)
{
strcpy(orig_options->loglevel, new_options.loglevel);
log_config_changed = true;
}
if (log_config_changed == true)
{
log_notice(_("restarting logging with changed parameters\n"));
logger_shutdown();
logger_init(orig_options, progname());
}
if (config_changed == true)
{
log_notice(_("configuration file reloaded with changed parameters\n"));
}
/*
* if logging configuration changed, don't say the configuration didn't
* change, as it clearly has.
*/
else if (log_config_changed == false)
{
log_info(_("configuration has not changed\n"));
}
return config_changed;
}
bool
reload_config(char *config_file, t_configuration_options * orig_options)
{
PGconn *conn;
t_configuration_options new_options;
/*
* Re-read the configuration file: repmgr.conf
*/
log_info(_("Reloading configuration file and updating repmgr tables\n"));
parse_config(config_file, &new_options);
if (new_options.node == -1)
{
log_warning(_("Cannot load new configuration, will keep current one.\n"));
return false;
}
if (strcmp(new_options.cluster_name, orig_options->cluster_name) != 0)
{
log_warning(_("Cannot change cluster name, will keep current configuration.\n"));
return false;
}
if (new_options.node != orig_options->node)
{
log_warning(_("Cannot change node number, will keep current configuration.\n"));
return false;
}
if (strcmp(new_options.node_name, orig_options->node_name) != 0)
{
log_warning(_("Cannot change standby name, will keep current configuration.\n"));
return false;
}
if (new_options.failover != MANUAL_FAILOVER && new_options.failover != AUTOMATIC_FAILOVER)
{
log_warning(_("New value for failover is not valid. Should be MANUAL or AUTOMATIC.\n"));
return false;
}
if (new_options.master_response_timeout <= 0)
{
log_warning(_("New value for master_response_timeout is not valid. Should be greater than zero.\n"));
return false;
}
if (new_options.reconnect_attempts < 0)
{
log_warning(_("New value for reconnect_attempts is not valid. Should be greater or equal than zero.\n"));
return false;
}
if (new_options.reconnect_intvl < 0)
{
log_warning(_("New value for reconnect_interval is not valid. Should be greater or equal than zero.\n"));
return false;
}
/* Test conninfo string */
conn = establish_db_connection(new_options.conninfo, false);
if (!conn || (PQstatus(conn) != CONNECTION_OK))
{
log_warning(_("conninfo string is not valid, will keep current configuration.\n"));
return false;
}
PQfinish(conn);
/* Configuration seems ok, will load new values */
strcpy(orig_options->cluster_name, new_options.cluster_name);
orig_options->node = new_options.node;
strcpy(orig_options->conninfo, new_options.conninfo);
orig_options->failover = new_options.failover;
orig_options->priority = new_options.priority;
strcpy(orig_options->node_name, new_options.node_name);
strcpy(orig_options->promote_command, new_options.promote_command);
strcpy(orig_options->follow_command, new_options.follow_command);
strcpy(orig_options->rsync_options, new_options.rsync_options);
strcpy(orig_options->ssh_options, new_options.ssh_options);
orig_options->master_response_timeout = new_options.master_response_timeout;
orig_options->reconnect_attempts = new_options.reconnect_attempts;
orig_options->reconnect_intvl = new_options.reconnect_intvl;
/*
* XXX These ones can change with a simple SIGHUP?
*
* strcpy (orig_options->loglevel, new_options.loglevel); strcpy
* (orig_options->logfacility, new_options.logfacility);
*
* logger_shutdown(); XXX do we have progname here ? logger_init(progname,
* orig_options.loglevel, orig_options.logfacility);
*/
return true;
}
